Declining AI Performance

Recent evaluations reveal a troubling decline in AI performance, particularly in math problem-solving and coding tasks. While the older version, GPT 3.5, shows improvement in basic math, the latest iteration struggles significantly, with only 10% success in coding challenges compared to 50% earlier this year. This decline has sparked debates within tech communities about whether the AI is genuinely worsening or if users are simply becoming more aware of its limitations.